CARRS Surveillance study: design and methods to assess burdens from multiple perspectives

This is a hybrid cohort-modelled cross-sectional multi-centre surveillance study to be conducted over a period of four years. Two cross-sectional surveys conducted three years apart on standalone representative samples of each of the three city-wide populations, using objective measures will permit estimation of the prevalence and trends of CMDs and their risk factors. Those enrolled in the first cross-sectional survey will be followed as a cohort in a three-year study to estimate (i) the incidence of new risk factors (such as obesity, hypertension, diabetes,); (ii) incidence of later-stage target organ diseases such as peripheral vascular disease, stroke, myocardial infarction, congestive heart failure, chronic stable angina, CKD, retinopathy, neuropathy, and amputation; (iii) assessment of health service utilisation and costs including hospitalisation and outpatient use and (iv) morbidity and mortality associated with CMDs.

The first cross-sectional survey has been completed with ongoing first year of cohort-follow-up. The survey was comprehensive, undertaking assessments of quality of Life (QoL), and socioeconomic burdens on individuals and families with regards to these diseases. Participants underwent anthropometric measurements, blood pressure (BP) assessment, and provided biochemical specimens. The cohort follow-up was limited to patient reports with recording of BP and anthropometry. CMDs and their complications were diagnosed using standard definitions and coded using the International Classification of Diseases 10 (ICD-10) codes.

The study sites are metropolitan urban settings with large, growing (due to continued births and migration from various parts of the country), and heterogeneous populations. Estimates suggest that population size in Chennai (4.68 million) [36], Karachi (13 million) [37], and Delhi (16.3 million) [36], and the diversity in their composition make these cities current and future archetypes of rapid socio-economic, demographic, epidemiologic, and nutrition/lifestyle transitions in the South-Asian region.

Utilising risk factor prevalence estimates from previously published Indian and Pakistani studies and anticipating a response rate of 80% with a design effect factor of 1.5 (to account for cluster sampling), the sample size estimates were generated for males and females in three age strata in each urban setting. As shown in table 1, the highest required sample size (3983 rounded-off to 4000 participants) permits each site to reliably estimate one or more of the CMD risk factors for each of the gender and age strata leading to a total sample size of 12,000.

With regards to the cohort follow-up, separate consent has been taken from participants to be followed up for three years or longer. An overall 15 - 25% loss-to-follow-up by the 3-year data collection time-period is anticipated due to the high probability of migration among the young population for job opportunities, marriage (in case of females), etc. Retention efforts (in the form of maintaining updated contact information; collecting contact details of friends and relatives; periodic reminder calls; courtesy calls/visits) have been put in place to keep track of participants and minimise loss-to-follow-up. Although the study at present is not powered to estimate incidence of CMDs and their risk factors, it has the potential to determine such incidence rates if the follow-up period is increased and the study is scaled up by adding follow up of subsequent cross-sectional samples.

Households were selected in each of the three cities using a multi-stage cluster random sampling technique. Each city has its own distinctive municipal sub-divisions, encompassing municipal corporations, wards and Census Enumeration Blocks (CEB), which were used sequentially as sampling frames to randomly select households. While wards were the primary sampling units (PSUs) for Chennai and Delhi, CEBs or clusters were the PSUs for Karachi. STATA version 10.1 (Statacorp, TX) [38] and data from the most recent census were used to randomly select the wards, CEBs, and households (defined below.). To give each household an equal chance of being selected for the study and to identify households constructed after the last census survey, manual listing and mapping of all households in each CEB was done before randomly selecting them.

Two participants, one male and one female, aged 20 years or older, were selected from each household based on inclusion and exclusion criteria given below. Two methods were used for within household sampling – (i) for households with one to two adults (≥20 years), the sampling strategy described in the 2002 Health Information National Trends Study (HINTS) in the USA was used [39]. According to HINTS, one or both individuals (one male and one female) were selected and enrolled into the study based on eligibility criteria and informed consent; (ii) for households with more than two eligible adults, the “Kish method” used in the WHO’s STEPS surveys [35] was applied. Recruitment of participants, and data and specimen collection were conducted through three visits to each participant’s place of residence, respectively (Visit-0, Visit-1, and Visit-2).

To provide consistency and reproducibility of the results across multiple sites, comprehensive and uniform data collection instruments were used to capture measurements (table 2). Household data were collected through interviewer administered paper questionnaires in English or the preferred local languages (Hindi, Tamil, and Urdu). Validated questions were derived from English questionnaires used in the WHO Multinational MONItoring of trends and determinants in CArdiovascular disease (MONICA) study [40], WHO STEPS studies [35], and from previous regional and national surveys. Using these, culturally-appropriate and methodologically-relevant closed-questions, an instrument for South-Asia was developed and pilot tested for face and construct validity prior to use in the study. Several sections of the baseline questionnaire (such as the QoL, CMD history, tobacco and alcohol consumption questions) were based on validated questionnaires that already exist in regional languages (Tamil, Hindi and Urdu). Questionnaires to elicit medical and treatment history, costs and QoL are being used to collect incident events during the ongoing cohort follow-up. Further, verbal autopsy is being performed using a reliable instrument to ascertain cause of death of participants who die during the course of follow-up and for whom either death certificate was unavailable or cause of death not certified. For the adapted instruments collecting a variety of CMD risks and diseases (e.g. tobacco, history of CMD, heart failure, and Chronic Obstructive Pulmonary Disease), the subjective history provided by participants was validated against laboratory and other diagnostic gold standards (e.g., salivary cotinine for tobacco consumption). These in-built steps to validate the self-reported data distinguishes the CARRS-Surveillance as a stronger model compared to the IDSP [31] and the INDEPTH network (International Network of field sites for continuous Demographic Evaluation of Populations and Their Health in developing countries (http://​www.​indepth-network.​org) [41].

Biological sample collection involved drawing 15 ml of blood (in fasting state) from each participant, collecting urine (early morning void), and 1000 to 2000 μl of saliva (fasting) in Salivettes. While blood and urine were collected from all participants, saliva was collected from 5% of the study participants (i.e. 200 participants per site). The samples were transported from field sites in cold chain to the laboratories for analysis. Sample aliquots were also stored in cryo-vials at - 80 degrees Celsius for future studies. The methods of analysis and external quality control have been standardised for all biological samples across the study sites (table 3). There is one exception in that Apolipoprotein A and B analyses were not conducted in Karachi due to lack of required laboratory facilities for the test.

Two clinical (BP and pulse rate) and eight anthropometric measurements of participants were taken during the visits: Clinical measurements - BP and Pulse rate. Anthropometric measurements - Mid-arm circumference, Waist circumference, Hip circumference, Triceps skin-fold, Sub-scapular skin-fold, Supra-patellar skin-fold, Height (Standing) and Body composition analysis by Bio-impedance.

The equipment and methods used for BP and anthropometric measurements were standardised and certified, and have been used in other epidemiological studies in the South-Asian population. BP was measured using electronic sphygmomanometer; Omron HEM-7080 and HEM-7080IT-E; Omron Corporation, Tokyo, Japan (certified by the British Hypertensive Society and the American association for Advancement of Medical Instrumentation [AAMI] protocols). Skinfold Calipers (Holtain Ltd., UK) and non-stretch measuring tape (Gulick II, Country Technology, Gays Mills, WI) were used to measure skin-fold thickness and body circumferences, respectively. Height was measured using a portable Stadiometer (SECA Model 213, SecaGmbh Co, Hamburg, Germany). Apart from these, body-composition analysers (instrument which measures body fat by sending out weak electric currents to measure impedance/electrical resistance by different tissues of the body); Tanita BC-418 in Delhi and Chennai, and BC-545 in Karachi were used to measure compartmental body fat distribution. To ensure standardisation, both instruments were tested in 50 male and 50 female participants to compare the parameters measured; i.e. weight, body mass index, basal metabolic rate, body fat and visceral fat. The results showed that all measured parameters were highly co-related for both males and females (r > 0.95, p < 0.05) between the two instruments, except body fat in males (r = 0.67, p = 0.67). Methods for BP measurement and anthropometric measures were based on the recommendation of the American Heart Association’s Council on High Blood Pressure Research [42] and the third National Health and Nutrition Examination Survey (NHANES-III) [43].

An online system was developed in an ‘open source’ platform PHP (Hypertext PreProcessor, scripting language for the web page/front end) and MySQL (My Structured Query Language) for data entry and database management at each site. This online database has been programmed to have automated in-built checks for logic which are ‘clinically reasonable’ (such as ranges, absolute and relative values, context and structure). It provides an efficient means of data entry, storage, and quality control. Data are available at the coordinating site for immediate feedback and timely corrections. The data have been stored in pass-word protected files and questionnaires in locked cabinets in all study sites, and only the study personnel have access to these. All information related to participant identification was de-linked from the data files before analysis to maintain anonymity.

Quality control (QC) strategies were applied using a framework which comprehensively considers each phase of the study and applies inter-related themes to every level of the study and are described in table 4. Apart from standardisation of laboratory methods (table 3), QC involved laboratory procedure assessment at two levels. Level-1, internal quality control: Local laboratories attached with the study centre followed their own internal quality control standard operating procedures (SOPs) to ensure accuracy, precision, and reproducibility. Level-2; external quality control: Irrespective of the nature of existing laboratory accreditation and / or SOP’s, all study site laboratories were enrolled into an external quality assessment program for clinical chemistry, HbA1c (glycated haemoglobin), lipid and human urine. This was implemented with support from the Randox International Quality Assurance Scheme (RIQAS), UK. The frequency of external quality control sample was two per month for clinical chemistry, lipid and urine, and one per month for HbA1c.

Apart from robust study methods and quality control mechanisms, the sample population recruited in our study conforms to the current age profile of the population in the two countries; about 65% of the population in India and 75% of population in Pakistan are below the age of 35 years [36, 54]. This demonstrates the success of the sampling strategy employed and has implications for the generalisability of findings. However, a limitation of the CARRS model is that the study setting is urban and does not include the larger rural population where the burden of CMD is also growing. An urban model, however, would anticipate the growth of urban areas the world over, and also provide insights into operational aspects of surveillance systems, and empirical evidence of successful implementation at lower costs. The New York City Health and Nutrition Examination Survey (NYC HANES) provides an aspirational model suggesting that surveillance in such metropolitan cities with diverse populations might provide a reasonable reflection of diverse and growing cities to each individual nation’s disease burdens [55].

Several LMICs have some structure to estimate the burden of NCDs, but a recent study by the WHO in 23 high burden countries (which includes India) showed that the existing systems are deficient in standardised data collection tools and often lack accuracy and quality [56]. The CARRS-Surveillance model addresses these technical standardisation and quality challenges in setting up national and regional CMD surveillance systems in South-Asia, but the task of scaling up will require political commitment, funds, and human resources. Although challenging, this is achievable and has been accomplished by a few LMICs. For example, the Ministry of Health in Jordan, in partnership with WHO and CDC, established the Jordan BRFSS in 2002 which conducts cross-sectional surveys every two to three years [45]. The national examination survey of NCD risk factors in Seychelles has been collecting data for planning and evaluating interventions since 1989 [46], and the Cuban system since 1995–96 [47]. Eleven Latin American countries (Argentina, Brazil, Chile, Colombia, Dominican Republic, Guatemala, Mexico, Panama, Peru, Uruguay and Venezuela) have new or emerging systems for serial national NCD and risk factor surveys in various stages of development [57]. These countries have demonstrated the utility of continuous surveillance in identifying high risk communities, planning interventions, and evaluating the effects of existing policies, thereby creating an evidence-base for steering national policies on NCD prevention and health promotion [45‐47, 57].